AI v3.1
Introduction
The AI v3.1 release enhances the functionality of the AI checklist by extending its ability to populate embedded fields inside Linked Record checklists and by introducing automation hooks that can invoke AI automatically. Previously, when users added a Linked Record checklist within a process, the AI checklist could not populate the embedded fields inside that linked checklist. With this release, the AI checklist can now fill those embedded fields, ensuring that data generated in one record is accurately reflected in the linked record checklist it references.
In addition, the release introduces automation hooks for AI, enabling admins and developers to write automations that can invoke the AI button automatically based on defined field events. This allows AI to execute without manual user interaction, making workflows more intelligent and context-responsive.
What’s included in this release
1. AI checklist support for embedded fields
The AI checklist can now populate Embedded Fields that exist within Linked Record checklists.
When a Linked Record checklist connects one record to another, the AI checklist in the parent record can now fill the Embedded Fields of the linked record automatically.
This ensures continuous data consistency between linked records without requiring duplicate entry.
This behavior is supported only when the AI checklist operates in standard mode. If AI Autofill is enabled, Linked Record checklists remain disabled, and Embedded Fields cannot be populated.
2. Automation hooks that can invoke AI automatically
The automation framework now supports the ability to invoke AI through automation hooks. This means admins can configure automations that automatically trigger the AI checklist when specific conditions or field events occur, eliminating the need for users to manually click the AI button.
Key capabilities
Automations can invoke AI using any event trigger recognized by the Unifize Budgie DSL, such as:
on-field-fill- when a field is filled or updatedon-field-update- when a field’s value changesor any other compatible trigger supported in the automation engine.
Using these hooks, an AI checklist can be executed immediately after an input field update or user action.
This allows Unifize workflows to incorporate AI-generated insights automatically as part of broader automation logic.
Integration details
AI invocation now participates in the same automation event structure as all other field actions in Unifize.
Each automation event carries system context, including field ID, chatroom ID, user, and timestamp, ensuring audit traceability and consistent execution.
These hooks can be combined with standard automation actions to:
auto-populate related checklist fields,
trigger dependent workflows when AI finishes processing, or
Notify users once the AI output is generated.
Access the full AI automation document here.
Before vs After
Scenario
Before AI v3.1
After AI v3.1
AI checklist populating Embedded Fields
The AI checklist could not populate Embedded Fields within a Linked Record checklist.
The AI checklist can now populate Embedded Fields inside Linked Record checklists directly from the parent record.
Automation invoking AI actions
Automations could not invoke or trigger the AI checklist field; AI could only be executed manually.
Automations can now invoke AI automatically when defined field events occur, allowing AI to run without manual input.
Process integration
AI execution was limited to user actions inside a single record.
AI can now be called through automations, allowing workflow-driven execution and tighter integration with process logic.
Best practices
Use standard AI checklist mode (not Autofill) when the checklist involves Linked Record or Embedded Fields.
Define clear and contextual prompts to ensure AI accurately fills the intended Embedded Fields.
Validate field type compatibility between parent and linked records before enabling AI filling.
When configuring automations to invoke AI, ensure that the trigger field and AI checklist belong to the same process context to avoid redundant calls.
Last updated